People and Projects

Intelligent feeder design for metal castings Elizabeth Jacob
with Roschen Sasikumar

Background : Casting simulation programs can test a given rigging design for defects but cannot generate a feeder design layout. Even with a simulation package, the initial design of the rigging system depends heavily on approximate methods and foundry experience. Simulation-based initial design can be developed by integrating the complementary strengths of a rigorous heat transfer analysis with practically derived rules for feeder design.

Methods : The solidification and temperature profile of any casting component is first simulated without feeders to identify the hot spots and the solidification pattern. Based on the solidification time data at every location of the casting, a novel clustering algorithm divides the casting into clusters around each hot spot. Each cluster is a feeding section requiring a separate feeder. From the volume and modulus of a feeding section, the size of the feeder is calculated. For every section, a genetic algorithm iteratively finds a class of optimal feeder dimensions given a user-specified feeder shape. The design can be iteratively refined using further simulations.

Results : The prediction of feeder locations and feeder optimized feeder dimensions by Virtual Feed software for a typical steel flywheel component is shown in adjacent figures. The hotspots in the cast component is shown in the figure. Based on the clustering algorithm, the user obtains the feeding locations. For the present case, Virtual feed predicts seven feeders and these feeders are designed using a casting software and the final figure shows the hotspot shifting out of the cast component